Prediction of Occult Hemorrhage in the Lower Body Negative Pressure Model: Initial Validation of Machine Learning Approaches.

This article focuses on evaluating multivariate machine learning approaches for early detection of occult hemorrhage (OH) using noninvasive physiological measurements in a simulated hemorrhage model known as lower body negative pressure (LBNP). Using two preexisting Mayo Clinic datasets, the study applied ensemble learning methods—random forest and gradient boosting—to compare the predictive performance of multivariate versus univariate models and invasive-plus-noninvasive versus noninvasive-only measurements. Results demonstrated that multivariate models significantly outperformed univariate ones, and noninvasive-only measurements predicted OH onset with accuracy comparable to models including invasive data, achieving clinically relevant prediction up to 10 minutes in advance. The study highlights the potential for noninvasive multivariate monitoring to improve early detection of hemodynamic instability, while noting limitations related to dataset scope and generalizability across clinical settings.